Our paper “A Robust Stochastic Approach to Mineral Hyperspectral Analysis for Geometallurgy” has been published in Minerals.

Collaborative work at the intersection of imaging and mining

This paper was born from a collaboration with Alvaro Egana’s team, bringing together expertise in hyperspectral imaging and geostatistical modelling. The challenge we addressed is fundamental to geometallurgy: how to reliably characterize mineral composition from hyperspectral data, accounting for the inherent uncertainty in both the measurements and the mineralogical interpretation.

Our approach introduces a stochastic formulation for mineral hyperspectral analysis. Rather than producing a single deterministic estimate of mineral abundance, we explicitly model the uncertainty through a probabilistic framework. This is paired with a hierarchical regression scheme that handles the complex, nested structure of spectral-to-mineral relationships.

Why stochastic matters

Deterministic approaches to mineral identification from hyperspectral images tend to be brittle – they give you a single answer with no indication of how confident you should be in it. In a mining context, where decisions about processing routes and ore blending depend on accurate mineralogical information, understanding the uncertainty is just as important as the estimate itself.

The hierarchical regression scheme allows the model to capture relationships at multiple scales, from broad spectral features down to subtle absorption bands that differentiate closely related minerals. This was technically challenging work, but the result is a more robust and honest characterization of mineral composition. I am grateful to the team for pushing this work forward – it opened doors to problems I continue to work on today.